求解广义动态非线性测量数据处理的信赖域法

Trust region method of resolving generalized dynamic nonlinear survey data processing

针对误差方程中同时包含非随机参数和随机参数的情况,提出了一种广义非线性动态最小二乘测量数据处理的新方法,即信赖域方法,并建立了相应的解算模型。计算中仅使用目标函数及其一阶导数信息,通过 BFGS 校正方法构造 Hessian 阵的近似,该算法不仅当迭代初值要求不严格时,仍保证收敛,而且在较强条件下具有二阶收敛结果。

Trust region method is proposed for generalized nonlinear dynamic least-square data processing aiming at error equations containing random and nonrandom parameters simultaneity, and corresponding algorithm model is built, where only function and its one-order derivative is used, and approximate Hessian matrix is built through adjusted BFGS. This algorithm not only can guarantee convergence when iterative initial value is not rigid, but also can guarantee second-order convergence in more rigid conditions.